Greger,
Do you have any idea how SER decides to include a "Route:" versus a "Record-Route:" header? If so, which piece of code in ser would write the second ACK below?
Here is a "200 OK" and two ACKs - The first ACK is good and the second ACK is bad because it should have a "Route:" header referring to the Sonus box.
100.99.99.99.99 is my SER proxy 100.10.10.10 is the public side of my firewall 216.50.50.50 is the ip of the Sonus box
So the ACK from SER to Sonus is incorrect.
Do you think this is worth posing to Jiri, Andrei, and company? All I know is that this ACK is bad when STUN is not used and it is good when STUN is used. I did upgrade my Grandstream, but that didn't help, and I've modified my nat_uac_test to use mode==19 rather than mode==3, but still get the same results.
Regards, Paul
U 2004/11/18 14:13:08.419098 100.99.99.99:5060 -> 100.10.10.10:5060 SIP/2.0 200 OK. Via: SIP/2.0/UDP 192.168.0.83;rport=5060;received=100.10.10.10;branch=z9hG4bKa70081ccdd52daf0. To: sip:14075551212@sip.mycompany.com;user=phone;tag=069c9797. From: "Paul (1002)" sip:9990010001@sip.mycompany.com;user=phone;tag=92691bb29380c100. Call-ID: e37c04be3e50ea72@192.168.0.83. CSeq: 21752 INVITE. Contact: sip:4075551212@216.50.50.50:5060. Record-Route: sip:216.50.50.50:5060;lr. Record-Route: sip:100.99.99.99;ftag=92691bb29380c100;lr=on. Accept: multipart/mixed, application/sdp, application/isup, application/dtmf, application/dtmf-relay. Allow: OPTIONS, INVITE, CANCEL, ACK, BYE, PRACK, INFO. Supported: timer. Content-Disposition: session;handling=required. Content-Type: application/sdp. Session-Expires: 240;refresher=uas. Content-Length: 244. . v=0. o=Sonus_UAC 18748 26881 IN IP4 216.229.118.76. s=SIP Media Capabilities. c=IN IP4 100.99.99.99. t=0 0. m=audio 35552 RTP/AVP 18 101. a=rtpmap:18 G729/8000. a=rtpmap:101 telephone-event/8000. a=fmtp:101 0-15. a=sendrecv. a=nortpproxy:yes.
# U 2004/11/18 14:13:08.428394 100.10.10.10:5060 -> 100.99.99.99:5060 ACK sip:4075551212@216.50.50.50:5060 SIP/2.0. Via: SIP/2.0/UDP 192.168.0.83;branch=z9hG4bKf4bb608e498ec61d. Route: sip:100.99.99.99;ftag=92691bb29380c100;lr=on. Route: sip:216.50.50.50:5060;lr. From: "Paul (1002)" sip:9990010001@sip.mycompany.com;user=phone;tag=92691bb29380c100. To: sip:14075551212@sip.mycompany.com;user=phone;tag=069c9797. Contact: sip:9990010001@192.168.0.83;user=phone. Call-ID: e37c04be3e50ea72@192.168.0.83. CSeq: 21752 ACK. User-Agent: Grandstream BT100 1.0.5.16. Max-Forwards: 70. Allow: INVITE,ACK,CANCEL,BYE,NOTIFY,REFER,OPTIONS,INFO,SUBSCRIBE. Content-Length: 0. .
# U 2004/11/18 14:13:08.429879 100.99.99.99:5060 -> 216.50.50.50:5060 ACK sip:216.50.50.50:5060;lr SIP/2.0. Via: SIP/2.0/UDP 100.99.99.99;branch=z9hG4bK2b35.552edb80cbf475b9be9ae3f9db23f960.0. Via: SIP/2.0/UDP 192.168.0.83;rport=5060;received=100.10.10.10;branch=z9hG4bKf4bb608e498ec61d. From: "Paul (1002)" sip:9990010001@sip.mycompany.com;user=phone;tag=92691bb29380c100. To: sip:14075551212@sip.mycompany.com;user=phone;tag=069c9797. Contact: sip:9990010001@100.10.10.10:5060;user=phone. Call-ID: e37c04be3e50ea72@192.168.0.83. CSeq: 21752 ACK. User-Agent: Grandstream BT100 1.0.5.16. Max-Forwards: 16. Allow: INVITE,ACK,CANCEL,BYE,NOTIFY,REFER,OPTIONS,INFO,SUBSCRIBE. Content-Length: 0. .
--- "Greger V. Teigre" greger@teigre.com wrote:
A few comments/questions after having looked at your config file:
- You are using 1.0.5.11 firmware on your Grandstream. 1.0.5.16 is the
latest and a lot of things have happened since 11. I would suggest you upgrade the firmware before you spend more time on it.
- The SIP conversation you sent earlier shows a conversation without using
STUN. In later messages you write about setting outbound and then stun. Have you tried setting outbound, but not stun? Any difference in Route?
- From 1.0.5.7, Grandstream started to send stun-detected ip and port when
symmetric NAT was found. My experience is that if stun is set on such a phone, you will not be able to catch the REGISTER with nat_uac_test("3"), but need to add the "rport different from src" test (16) found in the cvs (and then use ("19")). If you contrary to what I have seen are able to catch the Grandstream as behind NAT with nat_uac_test("3"), I would love to see the trace of that conversation...
- I don't have much experience with proxying, so this is a shot in the dark:
It looks like the ACKs get rewritten by ser because the ACK is part of a nat'ed conversation. Could fix_nated_contact() be doing this? I haven't had time to check the code yet. What would happen if you added a search on Route:
if (method=="REGISTER" || (!search("^Record-Route:")) || (!search("^Route:"))) ) {
fix_nated_contact();
The way I read the mesage from the Sonus guys, the ACK with Route in it should not be touched. Or am I wrong? However, I'm pretty sure that such a exception to fix_nated_contact is not common...
Well, my 2c worth... g-)
----- Original Message ----- From: "Java Rockx" javarockx@yahoo.com To: "Greger V. Teigre" greger@teigre.com; "ser users" serusers@lists.iptel.org Sent: Thursday, November 18, 2004 05:55 PM Subject: Re: [Serusers] Revisted Error: force_rtp_proxy2: can't extract bodyfrom the message
Thanks Greger!
Yes, I'd love another set of eyes. Attached is my entire ser.cfg (bogus IPs of course).
Regards Paul.
--- "Greger V. Teigre" greger@teigre.com wrote:
Paul, I'm not sure if this went on the list? I get digests...
Thanks for your answer on content-length, I'll try it out. I went back to your original message on ACK and Route. I cannot understand how ser should process an ACK with STUN differently from without. If this assumption is correct, you would probably have a different execution path in your ser.cfg dependent on STUN or not. It could have something to do with what I wrote in my last email with regards to the error in Grandstream when behind symmetric NAT. I use nat_uac_test("19") to catch Grandstreams with firmware 10.0.5.16 as the contact and via will be public_ip:portfromstun while the request comes from public_ip:portfromsiprequest.
I'm not confident I can help here as this is at the edge of my knowledge of the topic, but if you want an external look at your config, I'll be happy to have a look at it. g-)
Java Rockx wrote:
I am using
if(!(search("^Content-Length:\ 0")) {}in my ser.cfg and it seems to have eliminated all errors. Honestly, I still have yet to test this with inbound calls from our PSTN provider's Sonus equipment.
My bigger issue is why would SER add "Route:" headers correctly to ACK messages that flow from my SER proxy to their Sonus box only when my IP phone is configured to use STUN?
I am using nathelper and rtpproxy and everthing seems to work just fine inside or outside of our firewall and no IP phones seem to need STUN for SIP messages and RTP to play nice with our firewall or client's firewalls as as far as I can tell my ser.cfg is good.
So IMHO one of two things is happening;
- I have an error that I'm not aware of in my ser.cfg related to
NATed versus non-NATed UACs * nathelper/rtpproxy is not usable when SER is interacting with other SIP proxies and STUN must be used.
Has anyone ever gotten SER to talk with other SIP proxies using NATed clients?
Regards, Paul
--- "Greger V. Teigre" greger@teigre.com wrote:
Hi, I've been following this thread as I have experienced the same problems myself. When I get incoming calls (both from Cisco IP-PSTN gateway and from other SIP phones) to a Grandstream behind symmetric NAT, the messages you have noted can be seen in the log when hanging up.
I was not certain as to the conclusion you ended with. Do you use the filter: if (!(search("^Content-Length:\ 0")) { force_rtp_proxy(); };
to avoid the errors? I have been thinking about testing on method and not call force on BYE and ACK. Have you tried this?
I also saw your question on RFC compliance and the Sonus equipment: In order to make Grandstream phones register properly when using STUN behind symmetric NAT, I had to patch nathelper with the rport != port of received address check. (I use 0.8.14 and I guess you already have the patch with the development version). The reason is that Grandstream attempts to rewrite the address using STUN even though it correctly detects a symmetric NAT. I have seen that this was introduced in a new firmware not long ago (release notes). This pussles me as sources I have seen claims this to be invalid behavior (which seems correct to me).
Best regards, Greger
Java Rockx wrote:
Hi All.
I've hacked my ser.cfg but can someone comment on why I would be recieving a "200 OK" with a
The change I made to my onreply_route is below. The only thing I can see about these messages versus others is that the CSeq says "CSeq: {some digits} BYE" with "Content-Length: 0".
So for these messages I'm just not calling force_rtp_proxy().
I don't know if this is a symptom of my Grandstream BT100 only of if other ATAs or IP phones do this.
Regards, Paul
onreply_route[1] {
if (isflagset(2) && status =~ "(183)|2[0-9][0-9]") { fix_nated_contact(); if (!(search("^Content-Length:\ 0")) { force_rtp_proxy(); }; } else if (nat_uac_test("1")) { fix_nated_contact(); };}
--- Java Rockx javarockx@yahoo.com wrote:
> Hi all. > > I've got nathelper and rtpproxy working very well with my firewall. > However I do still recieve these messages in my syslog. I am only > catching 183 and 2xx errors in my onreply_route so I'm very > confused how to prevent these errors. > > I'm using ser-0.8.99-dev12. Can anyone give me some advise? > Cheers, > Paul > > NOTE: The SIP message that caused these errors is at the bottom of > this message. > > 0(27011) ERROR: extract_body: message body has length zero > 0(27011) ERROR: force_rtp_proxy2: can't extract body from the > message 0(27011) ERROR: on_reply processing failed > > My onreply_route is here:
=== message truncated ===
__________________________________ Do you Yahoo!? The all-new My Yahoo! - Get yours free! http://my.yahoo.com